Skip to main content
Premium Trial:

Request an Annual Quote

Johnson & Johnson Scientists Discuss Microarrays in Pharma Research



Dhammika Amaratunga

Johnson & Johnson Pharmaceutical research & development
Raritan, NJ

Jackson Wan

Johnson & Johnson Pharmaceutical research & development
San Diego, Calif.

Q What are the biggest issues you have to deal with in using microarrays for pharmaceutical R&D?

AWan: The biggest issue is probably normalizing data between chips. You have to apply multiple normalizations at different levels. You normalize between chips for the same sample done the same day, then between different samples, then between different animals (if you are doing rats), and between species, then by projects, then by day.

QWhat statistical method do you use to normalize data?

AWan: We use a smooth spline, a non-linear method. Genes expressed at very low levels tend to be normalized differently than genes expressed at high levels. So you need to normalize different parts of the expression levels different ways.

QWhat kinds of arrays do you use?

AAmaratunga: We have our own spotted cDNA arrays. We have just ordered a few Affymetrix chips.

QWhy did you switch to Affymetrix chips?

AWan: We didn’t use Affymetrix in the beginning because of the price. We’ve done 20,000 experiments in the last year, and to do them all on Affy would be too expensive. But recently they have given us a price break. While we like the cDNA technology, oligo chips are the way of the future.

QWhy are oligos better than cDNA arrays?

AWan: With cDNAs you are limited to the clones available, and there are a lot of variables to make consistent--growing the bacteria, doing miniprep, having to PCR many clones. Not too many people have got it down 100 percent. If you use Affy or another vendor you have a lot less room for error.

QWhat kinds of experiments do you do with microarrays?

AWan: One type is reference experiments, where basically the idea is to find out where a gene is expressed. The other type of experiment tells you what genes are expressed and where they are expressed. More theoretically, you take cancer vs. non-cancer controls, or treated vs. non-treated. A third type has to do with clinical trials. There are a lot of different tissues and samples from trials, and we can use microarrays to see whether gene expression patterns correlate with drug response.

QSo you are doing pharmacogenomics?

AWan: Exactly. We are trying to find correlations that we can apply later. The other kind of experiment is toxicology, testing compounds we know are toxic, non-toxic and looking at expression in multiple organs, then trying to find a pattern of expression that would predict toxicity. We have stored these patterns in databases to match patterns in toxicological experiments in the future.

QWhat kinds of software do you use to analyze microarray data?

AAmaratunga: We use quite a bit of different packages. I have been using S-Plus. But we also use the public domain software such as the Stanford offerings. We also collaborate with OmniViz, a company that uses visualization in its software package, GalaxyView. Without that kind of tool it is almost impossible to look at large amounts of data.

QWhat do you think is needed in microarray data analysis?

AAmaratunga: Even though you have a lot of data in microarrays you have relatively little data per gene. There is a bias that can creep in, and the findings you get may not be reproducible. One of the issues is how do you get information from other areas to beef up your findings. You can look at the other genes that are in the experiment and this will give you an idea of the precision variability of the experiment. Also, are there other experiments in which these same genes were run? You’d like to get information about the pathways and ancillary information about what’s known in the literature. One of the best ways we can get most information about microarrays in the near future is to try and consolidate all of these pieces of information.

The Scan

Study Links Genetic Risk for ADHD With Alzheimer's Disease

A higher polygenic risk score for attention-deficit/hyperactivity disorder is also linked to cognitive decline and Alzheimer's disease, a new study in Molecular Psychiatry finds.

Study Offers Insights Into Role of Structural Variants in Cancer

A new study in Nature using cell lines shows that structural variants can enable oncogene activation.

Computer Model Uses Genetics, Health Data to Predict Mental Disorders

A new model in JAMA Psychiatry finds combining genetic and health record data can predict a mental disorder diagnosis before one is made clinically.

Study Tracks Off-Target Gene Edits Linked to Epigenetic Features

Using machine learning, researchers characterize in BMC Genomics the potential off-target effects of 19 computed or experimentally determined epigenetic features during CRISPR-Cas9 editing.